risks, coping and roles of access to financial services · famine, borrow from kin, temporary...

Annals of the University of Petroşani, Economics, 16(1), 2016, 5-30 5

RISKS, COPING AND ROLES OF ACCESS TO FINANCIAL

SERVICES

MEHERUN AHMED, BAQUI KHALILY

*

ABSTRACT: Households move strategically to smooth consumption in the event of

economic shocks. This paper comprehensively analyzes various coping schemes adopted by

households in the event of different types and intensity of economic shocks. We conclusively find

that erosive coping is a dominant strategy for households except asset shocks. It is also evident

that the higher the total loss suffered the greater is the likelihood of adopting erosive coping

schemes for any of the three economic shocks. Contrary to the findings related to asset shocks,

the household is high likely to adopt erosive savings, help from relatives in case of economic

shocks. In addition to these two methods, households also takes up new loans and mortgaging

land when it encounters expenditure shocks. The househols is more prone to adopt multiple

strategies in case of income and expenditure shocks.

KEY WORDS: coping strategies, economics shocks, microcredit, saving,

consumption expenditure, well being.

JEL CLASSIFICATION: O12, Q54.

1. INTRODUCTION

The relationship between poverty and natural disasters in the developing world

has been a topic of interest and debate among the academics and the policy makers.

Households in developing countries face different types of shocks. Some are particular

to one household only. These are called idiosyncratic shocks. Again, some shocks, like

natural disasters affect the entire village, or a community or a trade or an occupational

group. These are referred to as systematic shocks. Households plan strategically to

smooth consumption in the event of income shocks followed by an exogenous natural

calamity. The set of coping strategies adopted by households depend on a number of

* Asian University for Women, [email protected]

University of Dhaka

mailto:[email protected]

6 Ahmed, A.; Khalily, B.

factors, especially, the types of crisis the households face and opportunities available to

them.

The people in the coastal belt of Bangladesh are yet to recover from the

devastation caused by Hurricane Sidr and Ila. Insurance against weather related risks

are beyond the reach of these people in the presence of an extremely thin insurance

market. Poor Bangladeshis, like citizens of most developing countries, are unable to

purchase various kinds of insurances, e.g., life, property or health insurances, from the

formal sector insurers as they are either unavailable or unobtainable.

Providing insurance to rural population in developing countries is quite

problematic because of asymmetric information and high transactions costs. Like credit

markets in poor regions, insurance markets are characterized by high transactions

costs, moral hazard, adverse selection, limited cash flows, low education levels of

clients, and weak enforcement mechanisms. Moral hazard and adverse selection remain

the two most important reasons hindering the formal insurers penetrating the rural

areas in the developing world. Moral hazard arises in the rural villages when farmers

when insured are less likely to adopt precautionary measures, or apply appropriate

amount of fertilizer, labor and other inputs, raising the chance of failure. Adverse

selection refers to the inclination of the relatively riskier farmer to purchase insurance.

Without much background information, insurance companies cannot distinguish

between riskier and safer clients, making it hard for them to maintain a profit margin.

In the face of these constraints the formal insurance providers (either the state

subsidized companies or the profit maximizing private-sector insurers) are uninterested

to provide services to rural populace even though the demand is rather high.

Recently in Bangladesh, microfinance organizations are playing an important

role to fill this vacuity by providing micro-insurance to their clients. Morduch (2004)

believed micro-insurance is going to be as successful as microcredit in the fight against

poverty. Not much is known about the demand side’s response to micro-insurance as a

precautionary coping mechanism in the face of any shock.

There has been a revolution in Bangladesh in terms of access to credit for the

poor people through the operations of various microfinance institutes. Extensive

research has been done evaluating the impact of microcredit in overall well being of

the households. Only a few studies investigated the role of microcredit in coping with

incomes shocks.

Using a new nationally representative dataset from Bangladesh, the broad

objective of this paper is to identify different coping mechanism adopted by affected

households in presence of a very thin insurance market and differential access to

formal and informal credit markets. Given the availability of some insurance to the

poor recently, and wide spread MFI operations, we also focus on the role of access to

credit and insurance in mitigating various income shocks, both exogenous and

endogenous.

2. LITERATURE REVIEW

Natural disasters affect the consumption pattern of households before and after

the event. Forward looking households in an effort to adopt risk mitigating techniques,

Risks, Coping and Roles of Access to Financial Services 7

incur ex-ante costs. Households also bear ex-post costs in coping with the aftermath of

natural disasters. Examples of such costs according to the literature, include loss of

uninsured assets, reduction in current consumption, liquidation of assets, interest paid

on loans from formal and informal sectors and the loss of human capital for the future

generation. The topic of risk coping and efficiency of the household has been

extensively researched. In this section I try to provide a review of some of the most

recent and relevant research pieces which is far from being exhaustive; rather

emphasizes the special research focus of this chapter. First, I try to present the various

coping mechanisms adopted by households for consumption smoothing purposes in the

event of an income shock as seen in the literature pertaining to the developing

countries. In the absence of formal insurance, and availability of credit, households

resort to various behavioural responses and also some informal arrangements.

Corbett (1988) classified the coping techniques into two broad categories:

precautionary and crisis strategies. Precautionary strategies are adopted in the wake of

repeated exposure to similar type of non-acute risks. In contrast, severe threat to food-

security forces households to resort to crisis strategies. In a similar study, Dunn and

Valdivia (1996) find that in the Andean semi-arid regions, wealthier households

owning more assets in the form of livestock, and therefore, are in advantageous

positions to adjust or mitigate the shocks ex-post, are less likely to adopt ex-ante risk

reducing strategies.

The most prominent Ex-Ante strategy adopted by households is to invest in

different income sources. As long as the sources of income do not co-vary perfectly,

risks to total income are reduced. Alderman and Paxson (1992) noted in their paper

that crop and field diversification, mix of farm and non farm occupations are quite

wide spread in the rural areas of developing countries. Morduch (1995) in his review

paper lists similar findings. Variability reducing inputs and production techniques are

often favored by households to smooth income. Households facing higher farm profit

volatility send members abroad for steady income flow. Rosenzweig and Binswanger

(1993) found that in India poorer farmers are more risk averse in the sense that they

adopt less risky production strategies. Farmers facing unpredictable environment,

select the blend of assets which are less sensitive to rainfall and generate low profit

levels.

Rosenzweig (1988), Urdy (1994) have found that households in the developing

world traditionally rely on social networks of extended family, friends and neighbors

and other informal institutions to mitigate the effect of the shock as Ex Post strategies.

They manage only partially to insure against shocks by engaging in informal credit

transactions and transfers. Fafchamps and Lund (2003) in a recent paper also find

similar results. More recently, in contrast to the African scenario, Morduch (2004)

identified several coping strategies for the households in Honduras after Hurricane

Mitch. In the presence of missing insurance markets, he found in his study using 1998

data that about 21% of the affected households drastically reduced consumption as a

main response to the hurricane. These households were unlikely to draw on insurance,

or erode assets, use savings or borrow funds.

It is well known that microcredit plays an important role in the lives of the

poor people in Bangladesh. Pitt and Khandkar (1998, 2002) find in their papers that


microcredit increases consumption and reduces poverty. It also helps smooth seasonal

consumption during the lean periods. Amin, Rai and Topa (2002) find that poor

households who participate in microcredit programs in Bangladesh tend to have

relatively better access to insurance and other consumption smoothing devices than

non-participants. Moreover, Rosenzweig (1988) found that access to financial

mechanisms such as credit and remittances enable the household to manage risks and

cope better.

Watts (1983) in his paper concluded that African households are forward

looking and their responses are not arbitrary. In his survey he listed the following

coping mechanisms in the order of frequency of adoption: storage of food during

famine, borrow from kin, temporary migration, sale of livestock, borrow from money

lenders, sale of domestic assets, sale of land and finally permanent migration. Cutler

(1986) also listed similar coping mechanisms in his study of Beja famine migrants in

Sudan. Pleitez-Chavez (2004) finds evidence that households that are subject to

adverse income shocks, tend to receive more transfers. He also found a positive

correlation between the magnitudes of the negative shock and the amount of transfers.

Yang and Choi (2007) found that in Philippines sixty percent of the exogenous

reductions in income is matched by remittance inflows from abroad. The authors find

evidence against the null hypothesis of unchanged consumption expenditures in

households with migrant workers but they found strong significant evidence of

variability in consumption expenditures in response to income shocks in households

without any migrant worker.

The other most prominent coping mechanism adopted by poor households in

response to shocks is accumulation or erosion of assets. In many parts of the

developing world poor credit-constrained households disproportionately hold

unproductive liquid assets as a precautionary measure. These precautionary reserves

take the form of livestock, foreign currency, durable goods, crop inventories, land etc.

(Udry 1995; Jalan and Ravallion, 2001; Gomez-Soto, 2007).

Even though the relationship between natural disasters and poverty is

extensively studied, there are still some gaps in this literature. There are only a few

studies investigating the household coping mechanisms in Bangladesh. The role of

micro credit and micro-insurance as a risk coping strategy is not well researched. This

is mainly because micro-insurance is a recent phenomenon and appropriate data are

virtually non-existent to the researchers for precise analysis. The role of overlapping of

loans, that is borrowing simultaneously from several NGOs as a coping strategy is also

not researched. None of the studies focused on the choice of different combinations of

strategies adopted by households depending on the nature of shocks. Using household

level data from a nationally representative survey conducted in 2010 that has a quite

rich, separate module on risk and coping strategies, it is possible to address these gaps

in the literature. Even though Bangladesh is a small country geographically, it is

visited by many natural disasters. The atrociousness of loss of lives and properties

reaches mammoth scale due to high population density. Thus this study bears

important policy relevance. The data-set also contains a whole list of demographic and

regional variables, allowing us to research the question with better accuracy and

statistical sophistication. In this paper, I try to address the following questions:


When individual households face economic shocks what type or combinations of

coping strategies do they adopt? Do choices vary by the intensity or types of

shocks faced?

Is micro-insurance a likely strategy adopted by households in the face of income

shocks?

What roles do micro-credit from Micro Finance Institutes (MFIs) play? Do the

households borrow from the formal or informal credit market? Or a mixture of

these three sources of credit?

How big a role do remittances play?

Do they dis-save? That is do they cope by eroding savings/assets/capital?

As mentioned before this study tries to bridge the existing gap in the literature.

There are only a handful of papers researching this important issue using Bangladeshi

data. Also most of these papers focus on a particular coping mechanism, e.g.

consumption reduction or migration or microcredit. This paper is comprehensive in the

sense that it analyzes all possible strategies for almost all types of disasters,

combination of all of these, and complimentarily and substitutability of strategies due

to the nature of various shocks.

A summary of the incidence of income shocks, both exogenous and

endogenous, by various household demographic characteristics and regional and

supply side characteristics is discussed first. A mean level comparison of the various

coping strategies is discussed in the next section. A comparison of various coping

schemes by income level, various demographic characters, nature and intensity of the

natural disasters etc, is also provided in the next section. A regression based analysis is

provided follows investigating the impact of shocks. Finally a discussion on the choice

of coping schemes is provided.

3. VARIOUS EXOGENOUS AND ENDOGENOUS SHOCKS IN

BANGLADESH: DEMOGRAPHIC CHARACTERISTICS OF THE AFFECTED

HOUSEHOLDS

Poor people in Bangladesh struggle to smooth consumption in the face of

various income shocks. Acute and chronic illness, loss of productive resources, loss of

livestock and fisheries, floods, droughts and other natural disasters, river erosion, fire,

crop failure, death of earning members etc. are some of the causes that affect family’s

income and consumption negatively.

The following table provides the summary statistics of households who are

affected by various types of shocks. It also shows the frequency distribution of affected

households in both rural and urban areas. In our sample, about 3.16 percent of the

households were affected by floods. A very small percentage of households reported

losses due to river erosion (0.42). 2.7 percent of the sample households suffered some

damage due to storms, cyclones or tornados. A very small number of households

reported losses due to fire, or loss in industry or sudden decline in remittance receipts.

Since majority of our sample households are in rural areas and predominantly

agricultural households, the data reveals that only a very small, 0.21 percent of the

households report any job loss or reduction in foreign remittances. The major shocks


that affected most of the households are loss of livestock and death and illness in the

family. About 6 percent of the sample households suffered some loss in income due to

death of livestock. 24 percent of the households report death or illness of adult earning

members of the households. The incidence of all types of shocks is higher in rural areas

than in the urban areas indicating that people in rural areas are more vulnerable.

Table 1. Percentage of household affected by different risk types and by region

Risk types

% of

Households

affected

Freq

% of

Households

affected in rural

Areas

% of

Households

affected in

urban Areas

Flood/excessive rain 3.16 262 3.75 0.85

Rain storm/Cyclone/tornado 2.66 222 3.14 0.77

River erosion 0.42 33 0.47 0.23

Catch fire 0.19 17 0.21 0.12

Suddenly lost service 0.21 18 0.14 0.51

Shortage of rain fall/drought 2.43 226 2.72 1.27

Crop disease 2.83 243 3.10 1.79

Cheated 1.22 109 1.23 1.20

Death of earning member 0.63 59 0.69 0.41

Family members are suffering diseases 23.34 2119 24.05 20.60

Accident of earning member 3.52 308 3.17 4.86

Accident of other member 2.65 261 2.59 2.90

Theft/robbery 2.99 275 2.74 3.97

Death of cattle 3.53 300 3.89 2.10

Death of poultry birds 5.66 504 6.11 3.89

Fish cultivation/loss asset 0.53 47 0.61 0.22

Loss in business 1.41 140 1.28 1.91

Litigation 1.66 149 1.80 1.12

Suddenly stop remittance from abroad 0.10 8 0.11 0.08

Others (specify) 1.89 172 2.11 1.01

Natural calamities like floods, cyclones, rain storms and droughts and cattle

diseases affect agricultural households more which are predominantly rural. We clearly

see a statistically significant difference between the percentages in the rural and urban

areas for these types of shocks. These are covariate shocks. Again there are other risks

which also affect the households or entrepreneur adversely but there are no significant

differences in acuteness of occurrences in urban and rural areas. Examples include

sudden loss of service, death of earning member, an accident of an earning member, an

accident of other members, theft/robbery, litigation etc. These shocks are individual-

specific and occur in isolation at different times to different individuals.

We probe this geographical difference more closely. In our data, about 47% of

the sample households report that they faced at least one crisis last year. Since weather

related shocks affect households that are predominantly dependent on agriculture, we

tested if there was a statistical difference between households that are mostly rural and

those that are urban.

As expected most of the disaster struck households are located in the rural area

and difference between the sample proportions of rural and urban affected households

are statistically significant at 5% level of significance or less.


52.8

40.56

47.2

59.44

0

10

20

30

40

50

60

70

Rural Urban

% of HHs affected % of HHs not affected

Figure 1. % of Affected and Unaffected Households by location

It is also found in the data that the geographical spread of the disaster or crisis-

struck households is more or less evenly distributed across 6 divisions of Bangladesh

except Rajshahi and Barishal. About 80% and 65% households respectively from these

two divisions were affected by some form crisis last years and the sample difference

between the affected and non-affected households in these two divisions are significant

at 1% or less implying the incidence of disasters disproportionately affected the

households in Rajshahi and Barisal division.

79.56

41.39

44.6837.69

65.34

41.87Barisal

Chittagong

Dhaka

Khulna

Rajshahi

Sylhet

Figure 2. Geographical distribution of affected households

It is suggested in literature that households solely dependent on agriculture

bear brunt of any natural calamity most. We saw indications of similar findings earlier.

These households try to insure themselves by diversifying crop pattern, land use

pattern etc. We tested the incidence of disasters by the occupation of the head of the

household. It turns out that within the agricultural households, the incidence is not

evenly distributed. Of the households whose head’s only income source is agriculture,

53.18 percent of them reported that they suffered from some shock and 46.49 were not

affected. A two-sample test of proportion with a z-value of 4.31 indicates that the

incidence is significantly different. 48.75 percent of the households that are not solely

dependent on agriculture and have alternative source of income along with agriculture


faced some income shock in the last three years and 50.38 percent reported no income

shock and the difference is statistically significant.

53.18

48.75

46.82

51.25

42

44

46

48

50

52

54

Agriculture Not agriculture


Figure 3. Percent of Affected and Unaffected Households by Occupation

This finding is consistent as the major shocks faced by households are

predominantly weather related. The significant share of the affected group also

reported loss of crops or livestock and death or illness of adult working members as

major shocks faced by them.

39.04

53.656.13

60.96

46.443.87

0

10

20

30

40

50

60

70

Landless Homestead only Agriculture land


Figure 4. Percentage of Affected and Unaffected Households by landholding

For the landless group there is a discernable difference in terms of shocks

suffered by the sample households. The households that only own homestead and the

households that own some agricultural or other land used for productive purposes

report significant loss of income due to shocks

About 54 percent of the households who only own homestead and 56 percent

of the households that own agricultural land face some form of crises in the last year.

These differences are statistically significant at 1% level of significance. But the

incidence of any income shock is less for the landless households. This is not


surprising as the landless households do not suffer from crop failures due to natural

calamities which are quite frequent in Bangladesh.

Figure 5. Incidence of shock by income groups

A very similar pattern is observed when we study incidence of shocks by

income status of the households. It is observed that in the sample, poorer households

face relatively fewer shocks compared to relatively richer households. About 52%

percent and 49 percent of the affected households are poor and non-poor respectively.

There is a significant difference in the sample proportions of affected and non-

affected households by their income levels. Poorer households report less income

shocks compared to non-poor households and the difference is significant at lower than

5 percent with absolute z-values of 3.43. The opposite pattern prevails for non-affected

households. For relatively well off households incidence of shocks are

disproportionately higher and is statistically significant at 5% or lower.

Table 2. Percentage of Households Facing Various Natural Shocks in the Last Year

According to Their Occupation and Participation Status in Micro-Credit Market

Shocks/Disasters

Participated in Microcredit

Programs

Occupation of the

Household Head

No loan Only one

loan

Multiples

Loans

Only

Agriculture

Other

occupation

Flood 8.98 9.73 3.43 6.61 10.03

Storm/cyclone/Tornado 6.75 3.43 1.59 4.92 4.20

Droughts 0.86 1.01 1.26 1.55 0.76

River Erosion 0.31 0.51 0.75 0.70 0.28

Loss of Crops 5.70 12.97 4.35 8.86 6.42

Loss of livestock 16.51 17.63 20.00 20.96 16.34

Loss in business 1.64 1.82 2.26 0.70 2.64

Fire 2.38 0.20 0.25 0.28 0.10

Death/Illness of Family Members 43.13 36.78 48.54 34.88 46.39

Loss of Jobs/Reduction in Remittances 0.55 0.41 0.17 0.98 0.24

Others 13.19 15.70 17.41 19.55 12.60

The following table describes participation in micro-credit programs and

occupation of the household head of different disaster affected households.

0

20

40

60

Affected Not affected Dont know

51.61 47.53

0.86

49.26 50.57

0.17

Non-poor

Poor


It is also not surprising to see that the households that suffer loss of livestock also

borrow from multiple MFIs. The same pattern is observed when families lose an

earning member. Loss of an earning member induces the household to borrow from

several sources and they also diversify occupation. Households in flood prone areas

also tend to have principal bread earner having other employment than agriculture.

They do it mostly as an insurance against natural calamities that would affect

agriculture most.

Thus we do observe through aggregate mean level data that households that are

relatively well off income wise, have relatively low level of human capital, and have

ownership of agricultural land, and predominantly rural are more prone to shocks or

crises.

4. VARIOUS COPINGS STRATEGIES ADOPTED BY HOUSEHOLDS IN

BANGLADESH

Risk affects households or entrepreneurs in different ways, with different

intensity and has a wider impact on the specific households' well-being. Depending on

the intensity and nature of risk households adhere to coping strategies for mitigating,

reducing and managing risk. The following table describes the frequency distribution

of various coping mechanisms adopted by our sample households.

Table 3. Households coping strategies (%)

Coping types

% of

Households

coped

Freq.

% of Households

coped in rural

Areas

% of Households

coped in urban

Areas

From saving 29.14 1331 25.90 45.68

From insurance 0.17 9 0.15 0.25

Help from relatives 10.00 473 10.20 8.99

Private help 0.80 36 0.89 0.33

Government help (VGD/VGF) 0.06 4 0.05 0.12

Government help (old allowance) 0.11 5 0.12 0.05

100 days program 0.04 2 0.03 0.10

Taken new loan 12.25 558 12.55 10.70

Sold labor in advanced 2.86 126 3.12 1.50

Sold crop in advanced 2.91 128 3.28 1.05

Sold land 1.36 60 1.52 0.54

Mortgaged land 2.00 87 2.29 0.53

No action taken 27.40 1208 28.56 21.48

Others (specify) 7.65 346 7.94 6.16

Erosive strategy remains the principle mode of coping when households face

any income shock in order to smooth consumption. Eroding savings leads the chart as

about 29.14 percent of the households have tried to mitigated shocks by drawing down

previous savings. It is not surprising to see that taking a new loan from a financial

institution is the second major coping scheme adopted by households. About 12.25

percent households cope by borrowing from a different MFI as the usual norm of a

MFI is to provide only one loan at a time to one individual. Again informal and formal

help/support is one of the major strategies for the mainly poor households who can’t


save or don’t have access to credit. In this regard informal sources such as friends and

family always play a major role. About 10.80 percent of households cope through

informal support. A very small fraction of the households relied on insurance or

government programs. Some other important strategies are selling labor and crop in

advance. About 6 percent of the sample households have adopted these two means.

Selling and mortgaging land is also a prominent way of coping /managing risk. About

3.5 percent of the households resorted to such strategies. Since land is a vital

productive resource, a household would be less inclined to sell the land. It would

There is a significant difference in the coping mechanisms adopted by

households based on the geographical location. Urban households adopt erosive

strategies in the sense that majority of the affected households about 46% to be precise

mitigate shocks through withdrawal of savings. The relevant number is 30% for rural

populace and the difference is statistically significant at 5% or less. Also strategies

such as advance sale of labor and crop, selling and mortgaging land are adopted mostly

by rural households and only a hand full of urban households and rural urban

difference is highly statistically significant. Since rural households are predominantly

agricultural and land is a chief mode of production, these are not surprising findings. In

terms of the other strategies, there are not much differences in the adoption frequencies

between the rural and urban populace.

5. COPING MECHANISMS OF VARIOUS DEMOGRAPHIC GROUPS

The following table provides a summary of different coping schemes

disaggregated by various demographic characteristics of our sample households. As it

is seen in the previous section erosion of savings, borrowing from relatives, advanced

sale of labor and crops, new MFI loans and mortgaging land are the favored options of

the households, we mostly focus on these coping strategies. It is important to know if

coping strategies vary by education, income, occupation and age of the household

head. This would help in designing appropriate policies and also help in targeting the

most vulnerable and needy group.

There is a positive association between income and erosion of savings as a

coping means. About 38% of the households belonging to the highest income quintile

cope by withdrawal of savings where as only 18% percent of the households in the

lowest income quintile adopt this mode. The difference is highly significant at 5% or

less. This finding is not surprising. Poorer households cannot save as much as the

higher income groups. Also we have seen earlier that poorer households are not as

affected by natural shocks as the rich households are. On the other hand informal

support (From friends and family) has a negative relation with the income that is low

income households mainly receive support from their kin when they are affected by

any income shock. Similarly new loan as a coping strategy is prominent for middle

income groups but not for the highest income quintile. However advance labor or crop

sale is also effective strategy for the low income households. Richest households in the

sample also sell the crops in advance to manage a risk.

Service holders mostly use savings or insurance to cope as 37.71 percent

service holders use this strategy. Taking a new loan is adopted by self-employed and


day laborers more compared to service holders. Advanced sale of labor would an

obvious choice for day-laborers and this is confirmed by data. The statistical difference

between various occupational groups in using advance sale of labor is highly

significant.

Table 4. Households characteristics of risk affected households

Characteris-

tics

Coping Strategies

Savings or

insurance

Relati

ves

GO and

NGO

supports

New

loan

Advanced

labor sale

Advanced

crop sale

Land

mortgage

or sale

No

action

taken

Inco

me

quin

tile

Lowest 18.69 13.98 0.41 10.07 3.69 3.02 3.28 31.70

2nd 26.41 13.03 0.38 13.46 3.99 2.85 3.56 27.02

3rd 30.74 10.46 0.11 14.21 3.82 2.70 2.10 25.79

4th 34.30 8.10 0.09 15.28 1.35 1.61 3.68 26.13

Highest 38.25 7.36 0.05 8.16 1.10 4.53 4.33 26.43

Occ

up

atio

n Service 37.71 11.78 0.07 8.56 1.15 2.53 2.45 19.52

Self

employment 29.94 8.03 0.21 12.00 1.67 3.49 4.46 31.05

Day labor 25.47 14.27 0.18 15.27 5.70 2.39 2.12 25.16

Others 28.48 10.90 0.63 8.21 1.28 2.13 3.16 27.52

Ed

uca

tio

n o

f

ho

use

ho

ld h

ead

Illiterate 24.49 12.76 0.39 13.27 4.29 2.83 3.78 28.32

Incomplete

primary 30.12 10.05 0.17 13.57 2.39 1.96 2.44 27.92

Incomplete SSC 34.66 9.35 0.00 9.49 1.47 3.26 2.65 24.79

Complete SSC 29.72 9.62 0.00 11.56 3.01 6.49 4.83 31.10

HSC 37.08 5.68 0.19 9.53 0.00 1.92 4.90 26.58

Above HSC 47.09 7.77 0.00 9.25 0.00 3.76 4.01 16.86

Ag

e o

f h

ou

seh

old

hea

d

Age below 30 30.40 11.00 0.08 11.73 3.19 2.09 3.05 26.13

Age between

30-40 30.80 9.76 0.11 13.45 3.99 2.87 2.84 26.24

Age between

40-50 29.41 10.53 0.09 12.41 2.07 3.00 3.56 27.53

Age between

50-60 28.94 11.11 0.58 10.70 1.79 3.39 2.59 30.93

Age greater

than 60 23.03 13.21 0.60 12.29 2.23 3.87 6.35 28.63

Adoption of various coping schemes varies significantly by the education of

the household head. There is a positive association between erosion of savings and

education level. Higher educated groups usually are service holders and belong to

richer income quintiles. They have relatively better access to savings and insurance and

data reveals that this is their preferred mode of mitigation. Another important result is

that less educated households rely on a new loan to cope compared to households

where the head as HSC or higher level of education.

Coping schemes do not vary significantly with age of the household head

except for the mode advanced sale of labor. Younger household head tend to adopt this

strategy more compared to relatively older age groups.

6. ROLE OF FINANCIAL INSTRUMENTS FOR COPING SHOCKS

Households cope in many ways; try to smooth consumption in the wake of

income shocks. The strategies adopted depend on the availability of the financial


instruments available to them. In literature, role of microcredit, remittances have been

documented as means of coping. Role of micro-insurance have not been studied

significantly. In this paper we try to find if access to these financial instruments enable

the households to cope better.

Table 5. % of Affected and Unaffected Households by Participation in Credit Programs

Shocks

Quasi-formal Formal Loan Informal Loan

Only One

Loan

Multiple

Loans

Only One

Loan

Multiple

Loans

Only One

Loan

Multiple

Loans

Affected Households 44.88 55.12 57.96 42.04 25.39 30.11

Not Affected Households 52.36 47.64 68.21 31.79 25.86 22.96

In Table 5 it is observed that most of the affected households have multiples

loans from microfinance institutes (MFIs) and their participation levels are

significantly different compared to non-affected households. The same trend is

observed for formal loans. They also borrow extensively from informal credit market

in the event of any shock or crisis. 30 percent of the affected households borrowed

multiple times from informal credit markets in last one year where as only 22 percent

of the unaffected households availed informal loans. These differences in participation

in credit market between the affected and non-affected households are highly

significant. In Bangladesh as revealed in aggregate level data, multiple borrowing from

formal, informal and quasi formal sectors probably is significant strategy choice by the

disaster affected households. The survey data reveals a curious pattern in terms of

remittance receipts by the affected households. It is seen that affected households

receive less remittances compared to non-affected households. This is contradictory to

the findings in literature about the positive association between remittance flow and

occurrence of any type of income shocks.

As Morduch (2004) points out, micro-insurance is going to be the next

revolution in terms of fight against poverty. Without access to any form of insurance,

the poor struggle to replenish their loss from any shock and are trapped in poverty

cycle for long. Role of micro-insurance has not been well documented in literature

mostly because of lack of proper data. In our survey, a very small percentage of

households have access to such financial instruments.

Figure 6. Remittance Receipts by Incidence of shocks

Affected

Not afected

40

45

50

55

Get remittance Donot get remittance

Affected

Not afected


Figure 7. Percentage of Affected and Unaffected Households by micro insurance

A significantly larger share of the affected households had micro-insurance

and the difference is highly statistically significant. 52 percent of the affected

households had insurance from the quasi formal sector compared to 46 percent of the

non-affected households. There is not much difference in terms of availing micro-

insurance depending on incidence of shocks.

Of the households that do not have any form of insurance, the incidence of

shock is evenly distributed among these households. Not surprisingly, a larger

proportion of the sample households that are affected by shocks have availed

insurance.

Even though the descriptive statistics gives us some indication of the coping

behavior of the households, it may be misleading as various forces can confound the

actual behavioral pattern. We try to precisely estimate the crisis coping behavior of

households through regression analysis when they incur some income shocks. It would

be interesting to see if the same pattern prevails in the regression analysis when

confounding and co-moving factors are controlled for.

7. REGRESSION ANALYSIS

Depending on the severity and the nature of the shocks, households adopt a

gamut of different strategies. They might also combine different strategies to guard

against transitory and permanent shocks. The questionnaire listed several possible

coping strategies (almost exhaustive) and also allowed the respondents to cite/mention

other ones not included in the list. The coping methods listed in the questionnaire are:

use of savings; insurance; financial help from relatives, NGOs, and government; new

micro-loan, mortgage or sale of land etc. The literature suggests that informal

insurance arrangements, borrowing from kin, community cooperatives etc may be

ineffective for shocks that are common to all members of the informal insurance

groups. Households also cope by borrowing from multiple sources, formal and

informal credit markets and MFIs. Remittances and sale of assets are also seen as

coping mechanisms adopted by households. There is not much known about the

simultaneous memberships of various MFIs, or combination of several techniques as

coping strategies in Bangladesh.

Micro insurance No

Micro insurance Yes

42 44 46 48 50

52

54

Affected

Households Not Affected


With this background information in mind, we proceed to identify for each

type of disaster or shock or crisis, the most likely coping method adopted by

households. The coping strategies might vary by the demographic and socio-economic

conditions of the affected households. We delve into that analysis with a view to

recommend and formulate appropriate, efficient policies, and to help in identifying the

right target groups etc.

In order to examine more rigorously the impact of natural disasters on

consumption expenditure, income and savings, we specify an empirical model which

permits tests of hypotheses concerning the type and severity of shocks, availability of

microcredit, erosion of savings and assets, migration of family members etc.

We basically interested in the following: What are the most likely strategies

adopted by households depending the nature and intensity of the shocks? Whether

having access to finance enabled the households to cope better in the event of an

income shock?

To assess the likelihood of various choice strategies adopted by households

based on the observed characteristics of the households and the nature of the income

shocks, we would adopt both bi-variate probit and multinomial conditional logit model

for our estimation.

In our data 14 coping strategies are listed. The multinomial logit response probabilities

of various coping strategies would be given by

)]exp(1/[)exp()|(1

j

h

hj xxxjyP

Where x is the vector of choice variables. The coping strategies, a random

variable y taken on values, J=1,…,14.

It is important that relative probabilities for the alternative coping strategies

depend only on the attributes of those strategies only, i.e., relatives odds between two

alternatives pass the Independence from Irrelevant Alternatives (IIA) assumption.

Given that individuals may simultaneously choose several of the coping strategies at

one point in time, it is clear that the response probabilities will not pass IIA test.

In order to tackle this problem we perform factor analysis. This process will

identify common coping capability of the households and reduce the number of 14

variables to a smaller number according common covariates. And these grouped

variables are most likely to be independent of each other. This is crucial for the IIA

assumption. Factor analysis is a statistical technique which explains a set of observed

variables in terms of a smaller number of latent variables called factors. These latent

factors are assumed to account for the correlations among observed variables. Thus the

common covariate of all these coping variables would capture the latent coping

capability of the affected households. I do not assume at the outset that one factor

would overwhelmingly explain the entire common covariance matrix of these 14

variables. On the contrary, I let the data determine the number of factors to be retained

and try to interpret them according to the factor loadings of the variables1. The

following figures show the results of the factor analysis in a nutshell.

1 We use factor analysis instead of principle component analysis as the latter imposes the restriction that

all the components completely explain the correlation structure among the variables. Factor analysis,


Figure 8. Scree Plot after Factor Analysis

Both the Kaiser-Guttman (only the eigenvalues that are greater than one) and

Scree plot2 (the curve levels off after the eigenvalue) suggest that we keep only 2

factors.

Figure 9. Factor Loadings after Factor Analysis

The factor-loading graph indicates that code 1 and code 3 are distinctly

different and the rest of the codes co-vary together. Code 13 which is no action taken is

separate than from the group. Code 1 is coping through savings and code 3 is help from

relatives. Thus our dependent variable would take j=4 values with “no action taken” as

base. Since the data on its own through factor analysis reflects that these four variables

are independent of each other, the IIA assumption would not be violated.

Estimating Equation. The model of coping scheme choice is given by:

accounts for the covariance of these variables in terms of a much smaller number of common covariates

(factors). Factor analysis does not force the common factors to explain the entire covariance matrix. That

is it allows the individual-variable specific influences to explain the remaining variances. 2 See appendix for the Scree plots for factor analysis.

.98

1

1.0

21

.04

1.0

6

Eig

enva

lues

0 5 10 15Number

Scree plot of eigenvalues after factor

code1code2

code3

code4code5code6code7code8code9code10code11code12

code13code14

-.5

0.5

1

Fact

or

2

-.5 0 .5 1Factor 1

Factor loadings


4,..1,0,)('

''

j

e

ejYprob

jj

jj

x

x

i

J takes 4 values in one model where we investigate the relative likelihood of

adoption In case of probit models, J takes only one value. X represents the vector of

control variables. We discuss the included controls and variables of interest in the

following section.

8. COMPARATIVE LIKELIHOOD OF ADOPTION OF VARIOUS

STRATEGIES

First we try to analyse the likelihood of adoption of the three strategies

comparing to no action to take depending on the type of shock or the intensity of

shocks by running multinomial logit models. The coefficients, even though difficult to

interpret, provide us with the direction of the likelihood and relative strength of each

choice. The four coping options are coping through eroding savings, help from

relatives, all of the others and the base is no action taken.

In addition to the standard household level demographic control variables like

family size, region of residence, age of the household head, some important household

level and supply side variables namely, education, electricity coverage, duration of

MFI membership, etc. are included in our regression analysis. Among the household

level characteristics, household head’s education level plays an important role in the

choice of coping strategies. Higher education implies access to information about

potential income shocks and available coping strategies. The household is able to make

better informed decisions regarding ex-ante and ex-post coping strategies when faced

with income shocks. A relatively poor household’s marginal disutility from income

loss is much higher than a wealthier household. Household’s permanent income level

would affect the choice of coping mechanisms. Education of household head and the

electricity coverage are used as proxies for household level permanent income. An

individual having a longer term relationship with MFIs would have more information

and more faith on the activities of the MFIs. It also reflects larger loan sizes which

enable the household to access bigger sums of money and confirms the bankability of

the client. Loan size is included to capture this effect. Rural areas are characterized by

a high degree of economics fragmentation. Long distances, difficult geography, lack of

paved roads, lack of public transportation make accessibility to markets difficult and

expensive. We would include divisional dummies and a binary indicator for rural area

to address the importance of regional and infrastructural facilities in the choice of

coping strategies.

The choice variables of interest are intensity of shocks, represented by only

one shock and two or more shocks last year; and types of shocks, natural, income loss

or death in the family etc. Model 1 studies the intensity and model two investigates the

types of shocks and their influence on households’ choice of shock mitigating schemes.

It seems from Table 6 model 1, that the number of shocks faced in the last one

year does not significantly increase or decrease the likelihood of choice of the three


coping methods. The log-odds of the choices to base outcome are not statistically

significant for the variables only one shock and two or more shocks.

In Table 6, model 2 we also observe that if the household faces a natural shock

it increases the likelihood of the log odds of the adopting the group of strategies as

compared to base. So is the case when there is a loss in business. It is interesting to

note that only in case of death and injury of adult working member in the family, the

household is significantly likely to adopt all possible options available to them as the

log odds between at no action taken and saving, help from relative and group strategies

are 1.9, 1.8 and 1.5 respectively. In case of such shocks, households would

significantly erode their saving as the odds ratio is higher than that of the other two

options. The likelihood of borrowing from relatives again is higher compared to the

group option of sale or mortgage of land; or advance sale of labor or crops, apply for a

new loan etc. This pattern is not observed for any other type of disasters faced by the

households. Erosion of savings or seeking help from kin are not most sought after

mechanisms in case of natural or business loss related shocks. One possibility of not

borrowing from kin may stem from the fact that natural disasters or bad crop or death

of live stock due epidemics like swine flu are systematic in nature, affecting a whole

region or community.

To test the consistency of the results we split the sample according to

ownership of land and occupation of the head of the household. The results are

presented in table 4 and table 5. A very similar trend as in Table 6 is observed. For the

households that own only homestead and the households that own some agricultural

land, any income shock through natural disasters, makes it more likely for these

households to adopt the group strategy to mitigate the shock as the log odds compared

to the base are positive. Any health injury or death in the family prompts the household

to adopt all 3 of the strategies compared to the base in both samples. The household is

relative more likely to erode savings, then it would borrow from kin and then adopt the

group option. Exact same pattern is seen when the sample is split by the occupation of

the head of the household.

The other control variables show expected signs in all the models. Households

that have electricity coverage erode savings or borrow from relatives. One year

increase in the education of the household head reduces the log odds of adopting any of

the strategies. Living in the urban area also reduces the likelihood of eroding saving or

seek help from relatives. All these variables are proxies for household’s income status

and rich households as we have seen in mean level analysis are better capable of

managing shocks and thus have advantageous positions of not to adopt erosive

strategies.

9. LIKELIHOOD OF ADOPTION OF DIFFERENT FINANCIAL

INSTRUMENTS

In order to identify how the households avail the financial instrument in case

of different types of disasters, we run several probit models where the dependent

variable is a binary indicator of borrowing from MFIs, purchase of micro-insurance,

purchase of formal insurance and remittance receipts. We focus on these variables as


the mean level analysis indicates the significance of these variables. The multinomial

regression analysis also shows that erosion of savings is more likely relative to other

options only in case of death and injury of a family member. Erosive strategies are not

significant in cases of other income shocks affecting the households. In the

multinomial regressions, the relative comparisons of choices were analyzed but we also

wanted to know individual likelihood of various financial instruments for policy

reasons. As factor analysis indicated particular grouping of the data, this type of

analysis was not possible in the multinomial framework because of IIA assumption.

In

Table 9 we find that households would borrow from MFIs if it faces any number of

shocks. The estimated coefficients for the binary variables, only one shock and two or

more shocks are statistically significant at 1%. Micro-credit appears as one of the

major instrument choice in coping against any type of shock. If the household is more

prone to various shocks, that is if it faces two or more shocks in a year, it is

significantly likely to borrow from MFIs. These households are also highly likely to

purchase micro-insurance or formal insurance as ex-ante coping mechanism.

Remittance doesn’t play a significant role and is not a significant choice for mitigating

the affects of shocks.

Households are positively likely to borrow from MFIs as an ex-post income

smoothing strategy if they face any natural disasters like floods, or droughts or

cyclones etc. The coefficient for loss in business or death of live stock is again positive

and significant implying the likely choice of micro-credit to help cope this type of

income shocks. If there is a death or injury or any other severe health problems of a

family member, households positively and significantly resort to loans from MFIs.

Natural calamities, business loss and theft or robbery etc. are positively

associated with the likelihood of purchasing of micro-insurance as an ex-anti coping

device. Formal insurance in Bangladesh is mostly life insurance but households are

entitled to borrow against their insurance amount. It is not surprising that there is a

positive and significant association between formal insurance purchase and loss in

business and theft and robbery.

There is an increased influx of remittances if the household suffers from theft

or robbery and if a family member dies or suffers from some injury. But there is no

significant association between remittance inflow and natural disasters or loss in

business and income loss due to damage or loss of crops or livestock controlling for

household demographic, income and region fixed effects.

10. CONCLUSION AND POLICY RECOMMENDATIONS

There is a dearth of literature analyzing the combinations of various pathways

by which households cope during a crisis in Bangladesh. To our knowledge this is first

study the investigated the relative likelihood various coping strategies for almost all

kinds of shocks faced by the households. It is quite comprehensive in that sense. The

study gives special attention to micro-insurance as being used as a coping strategy by

the poor segment of the society.


We observe that people who are tied to land, i.e., whose major occupation is

agriculture cope by adopting a group strategy involving new MFI loan, Sale and

mortgage of land and labor etc. to mitigate losses due to an exogenous natural disaster

like floods, cyclones, storms and river erosion etc. The other prominent strategies

adopted by these agricultural households are erosion of savings loans from relatives

when there is a death or sickness in the family.

A very similar pattern is observed in terms of adoption of various coping

schemes when agricultural households face any income shock (loss in crop, livestock,

industry etc.) or severe illness or death in the family. Borrowing from MFIs is a major

coping strategy when the household faces income shock from natural calamity or loss

in income and this pattern prevails irrespective of the household’s income and land

ownership status, occupation and education level of the household head.

In almost all of these scenarios, the role of loans from NGOs, Government

programs and loans from formal credit market seem negligible for all types of crises

and for all types households irrespective of their socio-economics characteristics.

Erosion of savings, loans from MFIs and help from relatives prominently top the list of

choices of coping strategies adopted by households. The significant role of microcredit,

micro-insurance in mitigating shocks has important policy implications. Death and

health shocks makes the household most vulnerable and forces it to erode savings and

borrow from relatives and also seek out other options. Increase in accessibility of life

insurance and other medical facility would prevent erosion of physical and financial

capital which is vital for productive efficiency of the households. Once diminished it is

often impossible for the poor households to replenish this capital and as a result they

might fall into poverty trap for good. Policy makers ought to pay attention to this as

well. This study also highlights the importance of health and life insurance and provide

insight for the Government to promote MFI operation in the insurance market as

formal sector insurance may be costly and inaccessible to the poorer segment of the

society.

11. TECHNICAL NOTE ON FACTOR ANALYSIS

In common factor analysis a small number of factors are extracted to account

for the inter-correlation among the measured variables. This helps to identify the latent

dimensions that explain most of the correlations among variables. We have a set of

bargaining measure variables, 1 ,.....,j Njx x . We want q common factors which

accounts for most of the covariance of the measured variables, Nx .

The standardized vector of observed variables can be expressed as a function

of correlation of variables and uniqueness associated with each variable.

x fA e

where, A=Nxq factor loading matrix represents the correlation coefficient s between N

variables and q factor factors.

The squared factor loading is the percent of variance in that variable explained

by the factor.


1f xq matrix of factors

e=1xN vector of uncorrelated errors with covariance equal to the uniqueness matrix,

, which is NxN diagonal matrix.

The variance of bargaining measures x, denoted by Z is composed into two

parts: z AA

The factor scores can be obtained by (regression scoring, Thomson 1951)

1f̂ A Z x

The scores are the indices that are estimates of components.

A very similar statistical procedure to factor analysis is PCA which accounts

for the maximum portion of the variance present in the original set of variables. PCA is

typically applied when the researcher instead of using all variables, wants to use some

indices that contain all the information present in the measures is the PCA which

derives a small number of components accounting for the variability found in a

relatively large number of variables. There are major differences between PCA and

FA. In FA, it is assumed that the variance of a single variable can be decomposed into

a common variance shared by all observed variables and a unique variance particular to

a variable. While in FA, only the common variance of the measured variables are taken

into account, Principle components are defined simply as a linear combinations of all

observed variables and PCA makes no distinction between common and unique

variance. PCA contains both common and unique variance.

Determining the number of factors in FA: The most commonly used

criteria in determining the optimal number of factors to be extracted are Kaiser-

Guttman rule and the scree test. The Kaiser-Guttman rule states that the number of

factors to be extracted should be equal to the number of factors having eigenvalues

(variance) greater than 1. A Scree plot illustrates the rate of change in the magnitude of

eigenvectors for the factors. The point where eigenvalues gradually levels off indicates

the maximum number of factors to be retained.


Table 6. Multinomial Logit Model: Dependent variable: Multiple coping strategies

Model-1 Model-2

Savings Relatives Others Savings Relatives Others

Affected by one shock -0.199 -0.690 -0.505

(0.828) (0.833) (0.714)

Affected by two or more shock 0.510 -0.494 0.465

(0.837) (0.847) (0.722)

Affected by natural shock? 0.018 -0.269 0.515***

(0.152) (0.187) (0.124)

Affected by livestock death? -0.158 -0.641** -0.120

(0.197) (0.265) (0.165)

Business loss or fish cultivation -0.198 -0.289 0.486**

(0.253) (0.321) (0.200)

Injury or death of family member 1.948*** 1.807*** 1.584***

(0.122) (0.141) (0.107)

Fraud or theft or rubbery -0.303 -0.553** 0.000

(0.184) (0.244) (0.155)

Irregularity in remittance Income -0.085 0.248 -0.079

(1.188) (1.170) (0.932)

note: *** p<0.01, ** p<0.05, * p<0.1. Each regression controls for age and education of the household head, Family size, Urban

Dummy and Division FE. No action taken as base

Table 7. Multinomial Logit Model of Multiple coping strategies, Homestead, Agri. Land

Homestead Agri. Land

Model-1 Model-2 Model-1 Model-2

Savings Relatives Others Savings Relatives Others Savings Relatives Others Savings Relatives Others

One shock -0.715 -0.565 -0.836 -0.304 11.715 -0.921

(0.921) (1.020) (0.824) (1.422) (553.008) (1.168)

2 more

shock 0.030 -0.372 0.106 0.329 11.671 -0.101

(0.930) (1.032) (0.832) (1.431) (553.008) (1.177)

Natural

shock -0.102 -0.377* 0.403*** -0.306 -0.460* 0.083

(0.163) (0.198) (0.133) (0.198) (0.265) (0.168)

Livestock -0.174 -0.630** -0.080 -0.181 -0.225 0.006

(0.209) (0.275) (0.174) (0.269) (0.360) (0.226)

Business

loss -0.153 -0.320 0.447** -0.513 -0.384 -0.083

(0.276) (0.350) (0.222) (0.397) (0.506) (0.330)

Injury or

death 1.747*** 1.642*** 1.421*** 1.930*** 1.954*** 1.586***

(0.131) (0.152) (0.115) (0.178) (0.227) (0.162)

Theft -0.293 -

0.739*** -0.045 -0.671** -0.761* -0.174

(0.198) (0.275) (0.167) (0.281) (0.417) (0.229)

Remittance -12.637 0.110 -0.833 1.065 14.990 0.847

(523.187) (1.175) (1.180) (1,019.898) (734.644) (949.005)

note: *** p<0.01, ** p<0.05, * p<0.1 . Each regression controls for age and education of the household head, Family size, Urban Dummy and Division FE.


Table 8. Multiple coping strategies, Agri. Household, Non-agri household

Agri. Household Non-agri household

Model-1 Model-2 Model-1 Model-2

Savings Relatives Others Savings Relatives Others Savings Relatives Others Savings Relatives Others

One shock -13.481 -12.967 -12.587 0.209 -0.483 -0.393

(660.047) (660.047) (660.047) (0.928) (0.931) (0.739)

2 or more

shock -12.917 -12.501 -11.455 0.930 -0.367 0.513

(660.047) (660.047) (660.047) (0.937) (0.947) (0.749)

Natural

shock -0.166 -0.064 0.907*** 0.046 -0.320 0.376**

(0.369) (0.363) (0.247) (0.169) (0.220) (0.146)

Livestock

death? -0.465 -0.562 0.140 -0.113 -0.636** -0.227

(0.506) (0.542) (0.319) (0.217) (0.305) (0.197)

Business

loss -13.954 -0.372 0.871 -0.161 -0.313 0.443**

(682.348) (0.909) (0.550) (0.261) (0.346) (0.217)

Injury or

death 2.167*** 1.977*** 1.593*** 1.905*** 1.803*** 1.585***

(0.295) (0.300) (0.224) (0.135) (0.163) (0.122)

Fraud,

theft, -0.741 -0.565 0.164 -0.271 -0.530** -0.047

(0.608) (0.679) (0.358) (0.196) (0.264) (0.173)


Table 9. The likelihood of taking different coping strategies for different shocks

Micro credit Micro-Insurance Formal Insurance Remittance

Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

One shock 0.138*** -0.022 -0.018 -0.115***

(0.030) (0.047) (0.035) (0.043)

2 or more shock 0.159*** 0.140* 0.169*** -0.054

(0.052) (0.076) (0.059) (0.077)

Natural shock 0.140** 0.185* -0.053 0.087

(0.060) (0.099) (0.071) (0.098)

Livestock death? -0.048 0.109 0.175* -0.025

(0.080) (0.117) (0.090) (0.130)

Business loss 0.208** 0.430*** 0.308*** -0.249

(0.099) (0.127) (0.105) (0.181)

Injury or death 0.122*** 0.026 -0.050 0.276***

(0.047) (0.072) (0.055) (0.076)

Fraud, theft, 0.003 0.196* 0.133 0.395***

(0.075) (0.107) (0.084) (0.107)

note: *** p<0.01, ** p<0.05, * p<0.1 . Each regression controls for age and education of the household head, Family size, Urban Dummy

and Division FE.


Table 10. Impact of shock on Food expenditure by microcredit use


With micro-credit with no micro-credit with micro-credit with no micro-credit

Model-1 Model-2 Model-1 Model-2 Model-1 Model-2 Model-1 Model-2

One shock 504.251*** 242.284 283.203** 246.470**

(155.418) (152.784) (117.839) (99.911)

2 or more

shock 696.328*** 468.967* 303.398 648.997***

(242.405) (272.932) (194.591) (189.083)

Natural

shock? -4.036 -384.074

-

883.657*** -430.979**

(263.898) (284.612) (228.404) (216.803)

Livestock

death? 725.370** 134.913 -195.086 205.834

(340.903) (431.629) (302.302) (280.472)

Business

loss -174.493 -1,434.503* 387.560 852.302**

(644.831) (764.027) (314.400) (348.046)

Injury or

death 136.025 -477.498** -354.725**

-

498.045***

(221.499) (238.929) (170.418) (170.082)

Theft or

rubbery 121.986 889.013* -194.845 761.782***

(407.839) (464.140) (258.043) (256.955)


and Division FE.

Table 11. Food expenditure by Homestead, Agri. Land with micro-credit, with no micro-

credit


with micro-credit with no micro-credit with micro-credit with no micro-credit


One shock 220.576** 141.423 467.578*** 336.946**

(108.084) (100.253) (173.619) (138.419)

2 or more

shock 354.375** 553.903*** 548.437** 752.077***

(173.686) (176.215) (264.985) (221.489)

Natural -

680.457*** -378.594*

1,065.859*

** -543.984**

(194.417) (194.924) (289.409) (238.668)

Livestock -57.952 37.823 -61.656 -173.189

(257.533) (253.897) (391.079) (312.227)

Business

loss 589.677* 855.410** 978.173*

1,638.940*

**

(304.159) (353.233) (516.727) (480.070)

Injury or

death -216.137

-

454.704*** -199.627 -410.632**

(152.040) (156.707) (244.785) (203.016)

Theft or

rubbery -201.393 659.383*** -469.686 254.912

(242.105) (246.500) (402.164) (325.811)


and Division FE.


Table 12. Loss remedy by Agri., Non-agri household with and without micro-credit


With micro-credit with no micro-credit with micro-credit with no micro-credit


One

shock 2,697.078*** 8,057.065*** 5,117.719*** 8,057.065***

(804.834) (1,125.787) (817.984) (1,125.787)

2 or more

shock 4,231.747*** 17,706.482*** 10,031.022*** 17,706.482***

(1,255.301) (2,130.583) (1,350.758) (2,130.583)

Natural 119.716 1,522.075 220.726 5,374.548

(1,739.264) (3,449.343) (2,035.910) (3,714.220)

Livestock

death? 2,136.594 6,617.952 7,807.381*** 727.303

(2,246.779) (5,231.110) (2,694.610) (4,804.967)

Business

loss -1,306.850 1,945.105 7,312.660*** 1,822.241

(4,249.864) (9,259.605) (2,802.442) (5,962.643)

Injury or

death -1,206.346 -1,328.950 -2,571.174* -7,049.298**

(1,459.828) (2,895.692) (1,519.045) (2,913.803)

Theft or

rubbery 8,510.063*** 8,308.617 9,047.960*** 14,205.101***

(2,687.932) (5,625.129) (2,300.097) (4,402.094)


Table 13. Loss remedy by Homestead, Agri. Land with and without micro-credit


with micro-credit with no micro-credit with micro-credit with no micro-credit


One shock 4,539.752*** 7,098.612*** 5,640.042*** 8,499.175***

(776.758) (1,080.437) (1,151.320) (1,825.205)

2 or more

shock 9,350.977*** 15,953.017*** 8,770.652*** 19,224.834***

(1,248.213) (1,899.091) (1,757.200) (2,920.582)

Natural 499.924 6,115.307** 865.478 5,265.253

(1,763.100) (3,115.941) (2,395.138) (4,359.311)

Livestock

death 5,425.930** 2,618.487 8,182.312** 2,634.572

(2,335.485) (4,058.662) (3,236.556) (5,702.858)

Business

loss 8,203.222*** 3,808.872 3,107.198 1,844.108

(2,758.313) (5,646.585) (4,276.412) (8,768.535)

Injury or

death -2,013.070 -3,175.750 -1,646.180 -2,588.628

(1,378.803) (2,505.026) (2,025.831) (3,708.120)

Theft or

rubbery 10,179.196*** 18,157.230*** 20,804.031*** 28,984.488***

(2,195.574) (3,940.414) (3,328.296) (5,950.974)

Remittance

income -9,985.892 18,011.999 13,535.173

(23,503.161) (20,698.010) (38,446.408)



REFERENCES:

[1]. Amin, S.; Rai, A.; Topa, G. (2003) Does Microcredit reach the poor and vulnerable?

Evidence from northern Bangladesh. Journal of Development Economics. 70(1), 59-82

[2]. Corbett, J. (1988. Famine and Household Coping Strategies. World Development, 16(9)

[3]. Fafchamps, M.; Lund, S. (2003) Risk Sharing Networks in Rural Philippines. Journal of

Development Economics. 71(1), 261-87

[4]. Morduch, J. (2004) Micro-insurance: the next revolution? Forthcoming in What Have We

Learned About Poverty? Edited by Abhijit Banerjee, Roland Benabou, and Dilip

Mookherjee, Oxford University Press

[5]. Pitt, M.M.; Khandker, S.R. (1988) The Impact of Group Based Credit Programs on Poor

Households in Bangladesh: Does Gender of the Participants Matter? Journal of Political

Economy. 106(5). 958-996

[6]. Binswanger, H.P.; Rosenzweig, M.R. (1986) Behavioral and Material Determinants of

Production Relations in Agriculture. Journal of Development Studies 22(3), 503-539.

[7]. Rosenzweig, M. (1988) Risk, Implicit Contracts and the Family in Rural Areas of Low

Income Countries. The Economic Journal. 98(393)

[8]. Udry, C. (1994) Risk and Insurance in a Rural Credit Market: An Empirical Investigation

in Northern Nigeria. Review of Economic Studies. 61(3), 495-526

risks, coping and roles of access to financial services · famine, borrow from kin, temporary...

Documents